Monitoring and Modeling of Saline-Sodic Vertisol Reclamation by Echinochloa stagnina

: Soil salinity due to irrigation is a major constraint to agriculture, particularly in arid and semi-arid zones, due to water scarcity and high evaporation rates. Reducing salinity is a fundamental objective for protecting the soil and supporting agricultural production. The present study aimed to empirically measure and simulate with a model, the reduction in soil salinity in a Vertisol by the cultivation and irrigation of Echinochloa stagnina . Laboratory soil column experiments were conducted to test three treatments: (i) ponded bare soil without crops, (ii) ponded soil cultivated with E. stagnina in two successive cropping seasons and (iii) ponded soil permanently cultivated with E. stagnina with a staggered harvest. After 11 months of E. stagnina growth, the electrical conductivity of soil saturated paste (ECe) decreased by 79–88% in the topsoil layer (0–8 cm) in both soils cultivated with E. stagnina and in bare soil. In contrast, in the deepest soil layer (18–25 cm), the ECe decreased more in soil cultivated with E. stagnina (41–83%) than in bare soil (32–58%). Salt stocks, which were initially similar in the columns, decreased more in soil cultivated with E. stagnina (65–87%) than in bare soil (34–45%). The simulation model Hydrus-1D was used to predict the general trends in soil salinity and compare them to measurements. Both the measurements and model predictions highlighted the contrast between the two cropping seasons: soil salinity decreased slowly during the ﬁrst cropping season and rapidly during the second cropping season following the intercropping season. Our results also suggested that planting E. stagnina was a promising option for controlling the salinity of saline-sodic Vertisols.


Introduction
Irrigated land degradation by salinization is a major constraint for agricultural production on irrigated soils, particularly in arid and semi-arid zones. Most of the reduction in crop yield is caused by salt accumulation in soil [1]. It has been estimated that 33% of irrigated agricultural land worldwide is afflicted by high salinity for various reasons, including low precipitation, high surface evaporation, weathering of native rocks, irrigation with saline water, and unsuitable cultural practices [2]. Clayey soils, such as vertic soils in the Niger River valley, which are irrigated for rice production in Niger, are particularly exposed to salinization because of their low hydraulic conductivity at water saturation which inhibits salt leaching [3,4]. The salinization process of vertic soils has been observed in irrigated perimeters of the Niger River valley and leads to the abandonment of agricultural land [5,6].
Economic and environmental challenges of irrigated land salinity have led to the development of several approaches to promote soil conservation and limit soil salinity. Many engineering-based remediation techniques have been developed to reclaim salt-affected soils, including physical, chemical, and biological remediation [7]. The last technique includes approaches such as phytodesalinization, which uses salt-tolerant plants or halophytes to decrease soil salinity and thus increase crop production [8][9][10][11]. This approach is increasingly investigated, both in field studies [12][13][14][15] and under controlled laboratory conditions [10,[16][17][18] since water ponding is difficult to apply in water-limited areas and less effective in fine-textured soils [4,19,20].
The processes underlying phytodesalinization have been reported by several authors [11,21] and include (i) the improvement of soil structure by root expansion, which promotes salt leaching; (ii) salt export in plant biomass, (iii) an increase in CO 2 partial pressure within the root zone and (iv) root proton release (for N2-fixing plants). However, the relative importance of these processes, particularly salt leaching and salt accumulation in plant biomass, has not yet been addressed well. Moreover, for Vertisols, phytodesalinization is influenced by shrink-swell processes, which govern water flow and solute transfer. Furthermore, the high clay content of Vertisols may limit plant root development [22].
Many studies have used simulation models to investigate water flow and salt dynamics in clay soils [23][24][25][26][27]. Few, however, have simulated salt dynamics in salt-affected soils undergoing remediation techniques. The Hydrus-1D model has been used to predict the effects of gypsum amendments on increasing the hydraulic conductivity and decreasing the salinity of clay saline-sodic soils [28,29]. Hydrus-1D has also been coupled with the UNSATCHEM module to simulate the Ca, Na, Mg and K contents of sodic Vertisols reclaimed with gypsum amendments and irrigated with water of varying quality. However, most existing studies have not considered the potential effects of plants to predict the salt dynamics in soil-plant systems, the plant-root modification of soil structure to improve salt leaching, or their high capacity to accumulate salts in their biomass.
Echinochloa stagnina is a perennial semi-aquatic grass characterized by a C4 photosynthetic metabolism. It develops a dense fasciculated root system which can colonize the soil profile even at depth. This dense root system can improve soil structure, i.e., increase soil porosity to favor preferential water flows towards a greater depth [15]. It is a fodder crop characterized by a high productivity (20 to 40 t/ha of dry matter) and a high nutritional value [30,31]. The E. stagnina plant is able to grow in salt-affected land. Indeed, several authors have reported the high phytoreclamation potential of E. stagnina in saline-sodic clay soil in Egypt [12] or in sodic soil and saline soil in the Niger River valley [15,32].
The aim of this study was to evaluate and simulate by a modeling approach, the phytodesalinization process of irrigated saline-sodic Vertisols cultivated with E. stagnina. Based on soil column laboratory experiments and simulation modeling, we compared two treatments of ponded cultivated soil columns with a control treatment consisting of bare soil columns.

Soil Characteristics
Soil cores used for laboratory experiments were provided from irrigated paddy fields of Kollo (13 • 16 35.32 N, 2 • 21 31.81 E), located in the Niger River valley 50 km southeast of Niamey (Niger). The soil characteristics (Table 1) showed that the soil is an acid saline-sodic Vertisol [33] with a high clay content (69.4%), an acid pH (4.7), a high electrical conductivity of soil saturated paste (ECe = 15.7 dS/m), and a low saturated soil hydraulic conductivity (<2.8 × 10 −8 m/s) [4,15]. The high exchangeable sodium percentage (ESP = 18%) shows that the soil is sodic and could have a degraded structure, with soil settling and a low hydraulic conductivity at water saturation induced by clay dispersion.  [34]; CEC: cobaltihexamine cation exchange capacity; C: concentration of corresponding exchangeable cation.

Soil Column Sampling
Six undisturbed soil cores were sampled from the topsoil layer (0-25 cm) of an uncultivated plot in the irrigated paddy fields of Kollo, chosen because of their high salinity. The soil columns (16 cm in diameter and 25 cm in height) were collected at the soil field capacity by a conventional coring method using manual pressure and PVC pipes. Each soil column was installed in the laboratory and filled with two layers (Figure 1) according to [35]: (i) a reconstituted topsoil layer (from 0-19 cm) compacted to 1.4 g/cm 3 of the bulk density; (ii) an undisturbed soil layer (from 19-25 cm) in which the initial soil structure was conserved.   [34]; CEC: cobaltihexamine cation exchange capacity; C: concentration of corresponding exchangeable cation.

Soil Column Sampling
Six undisturbed soil cores were sampled from the topsoil layer (0-25 cm) of an uncultivated plot in the irrigated paddy fields of Kollo, chosen because of their high salinity. The soil columns (16 cm in diameter and 25 cm in height) were collected at the soil field capacity by a conventional coring method using manual pressure and PVC pipes. Each soil column was installed in the laboratory and filled with two layers (Figure 1) according to [35]: (i) a reconstituted topsoil layer (from 0-19 cm) compacted to 1.4 g/cm 3 of the bulk density; (ii) an undisturbed soil layer (from 19-25 cm) in which the initial soil structure was conserved.

Experimental Design
The experimental design was composed of 3 treatments (2 replicates each): (i) ponded bare soil without a crop was considered as a control (CT); (ii) ponded soil cultivated with E. stagnina (CEs) in two successive cropping seasons separated by a total harvest, and (iii) ponded soil permanently cultivated with E. stagnina (CEp) with a staggered harvest every 4 months. At 5, 15 and 21 cm of soil depth, each column was vertically equipped with a porous ceramic suction cup to extract the soil solution, and with a micro-tensiometer to measure the soil pressure head (Figure 1). The soil solution was collected every 2 weeks with the porous ceramic suction cups, and the pressure head was measured at each depth every 10 min and recorded with an Almemo 5690-2 data logger. The temporal variation in the mass of each soil column and its drainage water (collected at the bottom of the column) was monitored with two digital balances. These masses were also measured every 10 min and recorded with the same Almemo 5690-2 data logger ( Figure 1). Each column was set on a gravel layer to facilitate free drainage.

Crop Production
The experiment was performed for two cropping seasons according to the cutting period of E. stagnina biomass: 4.5 months for each season separated by a 2-month intercropping season.
At the beginning of the first cropping season, saturated soil columns of CEs and CEp were cultivated with E. stagnina plants by transplanting three cuttings (15-20 cm tall) per column. These were grown under controlled conditions: artificial light at 1300 lux (i.e., 1.08 kW/m 2 ) provided for 12 continuous hours per day by two horticultural lamps with 8 neon tubes of 54 W each, located 70 cm above the plants. Room temperature was maintained at 30 • C during the day and 20 • C at night by a radiator on a timer and a thermostat.
During the cropping season, soil irrigation twice per week was performed uniformly on all columns with non-saline water (EC of 0.3-0.4 dS/m), by adjustment of the water layer above the soil surface to 7 cm for each date. These conditions were similar to natural conditions of E. stagnina growth in the Niger River valley.
An inorganic NPK fertilizer (12-15-20) was applied twice per cropping season according to [35], with doses of 200 kg/ha applied one week after planting and of 150 kg/ha one month after the first application. At the end of each cropping season, the yields of aboveground crop biomass (fresh and dry biomass) were evaluated by a harvest of all the biomass on CEs columns and most of the biomass on Cep columns, leaving 10 cm of plant stem at the soil surface which allowed the crop to regenerate during the next season after water irrigation. After the intercropping season, the second cropping season was started by planting new E. stagnina cuttings on CEs columns.

Soil Sampling and Soil Salinity Monitoring
For the initial state, a composite soil sample was collected from the reconstituted soil of all columns and from the undisturbed soil layer of each column. At the end of the experiment, the columns were destroyed, and soil samples were collected from depths of 0-8, 8-18, and 18-25 cm.
The soil electrical conductivity was measured in a 1:5 soil/water extract (EC 1:5 ) for each soil sample according to ISO standard 11265 [36]. The soil electrical conductivity measured in the saturated paste (ECe, dS/m) was estimated as: with 5.8 as the conversion factor for clay soils [37]. The EC of the soil solution (ECw), collected with the porous ceramic suction cups, was measured every 2 weeks at 25 • C. The SS was calculated for the 0-8, 8-18 and 18-25 cm soil layers using Equations (2) and (3). First, the total amount of salt in a soil sample (TS, g) was calculated using Equation (2) according to [4] with the relationship: with M as the soil sample mass (g), BD as the soil bulk density (g/cm 3 ), and d as the thickness of the soil layer (cm).

Salt Accumulation in Plant Biomass and Salt Balance in the Soil
The dynamics of plant biomass on the columns (biom) were estimated according to the method in [35] which was used for a logistic model [38] to establish a relationship between the weekly measured plant height, initial biomass, and final biomass. This logistic model was applied to each column and independently for each cropping season using the SSlogis function of R software [39,40].
At the end of each cropping season, the salt amount in the aboveground plant biomass was measured by summing Ca, Na, Mg, K, Cl, P and S biomass concentrations, as reported by [41]. The concentrations of cations (Ca 2+ , Na + , Mg 2+, K + ), anions (Cl − , SO 4 2− , PO 4 2− et NO 3− ) and total sulfur (S) were measured in the aboveground plant biomass according to ISO/CEI standard 17025 using (i) inductively coupled plasma atomic emission spectroscopy for Ca, Mg, S, and K and (ii) atomic absorption spectrometry for the other elements.
The salt balance in the soil was estimated weekly from the amount of salt removed in drained water, the amount of salt accumulated in plant biomass, and the salt input from irrigation water. The salt content of drained water and irrigation water was measured weekly using the oven-evaporation method.

Desalinization during the Intercropping Season by Water Ponding
Water was periodically ponded during the intercropping season to remove salt concentrated on cracked soil. Doses of irrigation water were applied uniformly to all columns. The water was first applied slowly (1.25 mm/h for 4 h) to promote salt leaching from the prism faces by preferential flow and then in a larger dose (2.5-5 mm) until soil saturation, to promote salt dissolution in the soil solution and its leaching. This procedure was performed every 5-10 days for 30 days, with no irrigation between the two ponding cycles. Drained water was collected at the bottom of each column at the end of each ponding cycle to measure its EC and the amount of salt.

Modeling Approach of Vertisol Desalinization
Hydrus-1D was used to simulate the dynamics of the soil salinity of Vertisols under conditions of permanent water saturation during E. stagnina cropping seasons and under conditions of non-saturation and cracked soil during the intercropping season. A bare soil irrigated at the same frequency as that of cultivated soil was also simulated.

General Model Description
Hydrus-1D [42] is widely used to simulate the dynamics of EC and SS in soil. It numerically solves the Richards equation (Equation (4)) for water flow and uses an advectiondispersion equation (Equation (5)) for solute transport in variably saturated porous media [39].
with θ as the soil volumetric water content (L3/L3), t as time (T), z as the vertical space coordinate (L), K as the hydraulic conductivity (L/T), h as the pressure head (L), and S as the sink term accounting for root water uptake (L 3 /L 3 /T).
∂θC ∂t with C as the solute concentration of the liquid phase (M/L), D as the dispersion coefficient (L2/T), and υ as the mean pore water velocity (L/T). The dispersion coefficient is defined as: where λ is dispersivity (L), which is considered a material constant independent of the flow rate. Since υ is obtained from predictions of the water flow model (water flux q divided by θ), λ is the only solute transport parameter needed to solve the convection-dispersion equation. It was fixed to a mean value of 0.15 cm following the recommendations of [43] for clay soils.

Soil Hydraulic Properties
Unsaturated soil hydraulic properties were described using van Genuchten-Mualem functional relationships [44,45]: with θ r and θ s as residual and saturated θ (L 3 /L 3 ), respectively; α (L −1 ) and n as empirical shape parameters; m = 1 − 1/n ; l as the pore connectivity parameter assumed to be 0.5 [41] and S e as the effective saturation. The parameters θ r , θ s , α, n and K s were estimated for each soil layer in each column for each season using pedotransfer functions predicted by the Rosetta model [46] based on the layer's particle size distribution (25.9% silt, 69.4% clay, 4.7% sand) and bulk density ( Table 2). The parameters α and n were adjusted by fitting measured retention curves according to [6]'s study on topsoil (0-40 cm) sampled on the same field. In HYDRUS, the time step is automatically adjusted between the minimum and maximum times steps specified by the user. In this study, they were 10 −5 and 5 days, respectively. Table 2. Soil hydraulic parameters used to simulate two soil layers during the first cropping season (S1), the intercropping season (IC) and the second cropping season in the six columns. The parameters are residual volumetric water content (θ r ), saturated volumetric water content (θ s ), a shape parameter (α ) that is related to the air entry potential of the soil, another shape parameter (n) and the saturated hydraulic conductivity (Ks). CT1 and CT2 are the columns with bare soil, CEs1 and CEs2 are the columns cultivated with seasonal E. stagnina and CEp1 and CEp2 are the columns cultivated permanently with E. stagnina.

Numerical Simulation
For the water flow models the initial conditions were defined using the measured head pressure distribution. The soil surface was subject to atmospheric boundary conditions with a surface water layer (maximum depth 7 cm) and specified values of irrigation and evaporation. Lower boundary conditions were defined as a seepage face (h = 0), which is usually applied to laboratory soil columns in which the bottoms are exposed to the atmosphere (gravity drainage of a finite soil column) [42]. Daily evaporation and transpiration were estimated from the changes in mass measured during the experiment, including the masses of irrigation water, drained water and plant biomass. For solute transport models, the upper and lower boundary conditions were defined respectively, by a concentration flux and zero concentration gradient. The initial conditions were defined by (i) the ECw of the soil solution measured at 5 and 21 cm for the simulation of EC dynamics and (ii) the initial SS calculated from the dry soil mass and initial salt content.
The EC and SS were simulated in three separate steps respectively using (i) a singleporosity model (SP) for water-saturated soil during the first cropping season (SP1), (ii) a double-porosity model during the intercropping season (DP) to represent preferential flow in cracked soil [47] and (iii) a single-porosity model during the second cropping season (SP2). SP was performed during the two cropping seasons because soil columns were permanently saturated by regular irrigation water, while DP was performed during the intercropping season because drying, causing cracked soil, presented preferential flows of water and solute salt was considered to be nonreactive in the soil [48,49].

Root Water Uptake and Salt Uptake
Hydrus-1D simulates root water uptake S using the approach of [50]: with α(h) as the water stress response function, which is a dimensionless function of the soil water pressure head h (cm) following Equation (12), and s P as the potential rate of water uptake [T −1 ].
with h 1 , h 2 , h 3 , and h 4 as threshold parameters. Water uptake is at the potential rate when the pressure head is between h 2 and h 3 , it decreases linearly when h > h 2 or h < h 3 , and becomes zero when h < h 4 or h > h 1 . We set h 1 , h 2 , h 3 , and h 4 to −10, −25, −200, and −8000 cm, respectively [50], values which are estimated as typical for fodder crops such as E. stagnina. The decrease in root water uptake due to salinity stress was described by the [51] function. A salinity threshold of 11.2 dS/m for EC and a slope of 3.8% for perennial ryegrass were selected from the function's database, based on the assumption that the parameters of E. stagnina were similar to those of perennial ryegrass.
We assumed that E. stagnina passively took up salt when its roots took up water [52]. Passive salt uptake was simulated by multiplying the root water uptake by the dissolved salt concentration, for concentrations below a predefined maximum concentration: All the dissolved salt is taken up by roots when C max is higher than C, while no salt is taken up when C max equals zero. The value of C max was defined by the salt content of plant biomass measured in the laboratory.

Statistical Analysis
Statistical analysis was used to evaluate the goodness of fit between laboratory measurements (O i ) and simulated values (S i ). The agreement between the predicted and observed data was evaluated with the root mean square error (RMSE) and Nash-Sutcliffe model efficiency coefficient (NSE): with O as the mean of observed values and N as the number of terms in the compared series.

Water Flow in Soil Columns
The changes in the soil pressure head on all the columns showed a saturated soil with a decreasing moisture gradient from the topsoil to the deepest soil layer during the cropping seasons (S1 and S2) and a desaturated soil with highly marked soil drying during the intercropping season (IC) (Figure 2). In general, the results of the HYDRUS model simulation reflected the same trends as the results measured with tensiometers, thus showing the efficiency of the model in simulating water flow in soil columns.

Electrical Conductivity of the Soil Saturated Paste and in the Soil Solution
The soil ECe was initially the same in the reconstituted soil layer at 0-8 and 8-18 cm in all columns (13.9 dS/m), but differed among columns in the undisturbed soil layer (18-25 cm) (15.3-21.2 dS/m) ( Table 3). At the end of the experiment, the initial ECe had decreased considerably in all soil layers in all columns. In the 0-8 cm soil layer the ECe decreased by 79-88% in all columns. In contrast, in deepest soil layer (18-25 cm), the ECe decreased less in bare soil (32-58%) than in soil cultivated with E. stagnina (72-83%), except in replicate CEp2, for which the ECe decreased by only 41%. However, in replicate CT1 a preferential water flow bypass was observed, especially at the beginning of the second cropping season due to soil cracking, causing an abnormally high transport of dissolved salt. So, we considered this replicate CT1 as nonrepresentative.
For each column, considering all depths (5, 15 and 21 cm) and seasons together, the measured and simulated ECw were in agreement, with a low RMSE (0.04-2.4 dS/m) and high NSE (0.60-0.99) (Figure 3, Table 4). Both the measured and simulated ECw decreased over time at all depths in all columns (Figure 3). The measured and simulated ECw differed most at a depth of 21 cm, particularly in CEp1 during the first cropping season, with an RMSE of 2.7 dS/m, and NSE of 0.45 (Table 4). The ECw decreased even more rapidly after the intercropping season at all soil depths in all columns. At the end of the experiment, the measured and simulated ECw were lower in columns cultivated with E. stagnina than in the bare soil replicate CT2 at all depths. The initial ECw measured at a 5 cm depth decreased by 76-88% in columns cultivated with E. stagnina and by 73% in the bare soil replicate CT2. At a depth of 21 cm, the initial ECw decreased by 54-92% in soil columns cultivated with E. stagnina, but by only 40% in the bare soil replicate CT2.

Electrical Conductivity of the Soil Saturated Paste and in the Soil Solution
The soil ECe was initially the same in the reconstituted soil layer at 0-8 and 8-18 cm in all columns (13.9 dS/m), but differed among columns in the undisturbed soil layer (18-25 cm) (15.3-21.2 dS/m) ( Table 3). At the end of the experiment, the initial ECe had decreased considerably in all soil layers in all columns. In the 0-8 cm soil layer the ECe decreased by 79-88% in all columns. In contrast, in deepest soil layer (18-25 cm), the ECe decreased less in bare soil (32-58%) than in soil cultivated with E. stagnina (72-83%), except in replicate CEp2, for which the ECe decreased by only 41%. However, in replicate CT1 a preferential water flow bypass was observed, especially at the beginning of the second cropping season due to soil cracking, causing an abnormally high transport of dissolved salt. So, we considered this replicate CT1 as nonrepresentative.

Salt Accumulation in Plant Biomass
The measured and simulated salt accumulation in dry plant biomass (shoot and root) were consistent (Figure 4), with a low RMSE (0.05-0.75 g) and high NSE (0.70-0.99). Moreover, the salt accumulation in plant biomass increased with plant growth in all cultivated columns: 4-6 and 8-11 g per column (equivalent to 440-660 and 880-1110 kg/ha) at the end of the first and second cropping seasons, respectively. Salt accumulation in the aboveground plant biomass represented 22-27% of the initial soil SS. However, Hydrus-1D predicted less salt accumulation in dry biomass than that measured, particularly during the second cropping season, in all columns except for replicate CEp2 during the first cropping season and the intercropping season.

Salt Accumulation in Plant Biomass
The measured and simulated salt accumulation in dry plant biomass (shoot and root) were consistent (Figure 4), with a low RMSE (0.05-0.75 g) and high NSE (0.70-0.99). Moreover, the salt accumulation in plant biomass increased with plant growth in all cultivated columns: 4-6 and 8-11 g per column (equivalent to 440-660 and 880-1110 kg/ha) at the end of the first and second cropping seasons, respectively. Salt accumulation in the aboveground plant biomass represented 22-27% of the initial soil SS. However, Hydrus-1D predicted less salt accumulation in dry biomass than that measured, particularly during the second cropping season, in all columns except for replicate CEp2 during the first cropping season and the intercropping season.

Soil Salt Stock
The simulated SS followed the same dynamics as ECe and agreed with the measurements from all columns ( Figure 5), with a low RMSE (0.3-4.0 g) and high NSE (0.70-0.99). Simulated and measured results showed that the soil SS, initially the same among treatments ( Figure 5), decreased for most of the experiment. At the end of the first cropping season, the SS had decreased by 5-14% in soil cultivated with E. stagnina (CEp and CEs), but had increased by 3% in bare soil (CT1 and CT2). The SS decreased more during the intercropping season, when water was periodically ponded on cracked soil. At the end of the intercropping season, the initial SS had decreased by 35-45% in soil cultivated with E. stagnina and by 14-22% in bare soil. At the end of the experiment, the SS was higher in bare soil (33-40 g per column) than in cultivated soils (9-22 g per column).

Soil Salt Stock
The simulated SS followed the same dynamics as ECe and agreed with the measurements from all columns ( Figure 5), with a low RMSE (0.3-4.0 g) and high NSE (0.70-0.99). Simulated and measured results showed that the soil SS, initially the same among treatments ( Figure 5), decreased for most of the experiment. At the end of the first cropping season, the SS had decreased by 5-14% in soil cultivated with E. stagnina (CEp and CEs), but had increased by 3% in bare soil (CT1 and CT2). The SS decreased more during the intercropping season, when water was periodically ponded on cracked soil. At the end of the intercropping season, the initial SS had decreased by 35-45% in soil cultivated with E. stagnina and by 14-22% in bare soil. At the end of the experiment, the SS was higher in bare soil (33-40 g per column) than in cultivated soils (9-22 g per column). The salt balance, calculated from the initial soil SS, salt inputs from irrigation water and the final soil SS at the end of the experiment, was positive in all columns, confirming soil desalinization (Table 5). Irrigation water contributed to 11-13% of the soil SS (i.e., 7.1-7.9 g per column) in columns cultivated with E. stagnina and to 6-8% of the soil SS (i.e., 3.4-4.4 g per column) in bare soil, due to the larger amount of irrigation water in columns cultivated with E. stagnina. At the end of the experiment, the SS had decreased more in soil cultivated with E. stagnina (by 65-87%) than in bare soil (by 34-45%).  The salt balance, calculated from the initial soil SS, salt inputs from irrigation water and the final soil SS at the end of the experiment, was positive in all columns, confirming soil desalinization (Table 5). Irrigation water contributed to 11-13% of the soil SS (i.e., 7.1-7.9 g per column) in columns cultivated with E. stagnina and to 6-8% of the soil SS (i.e., 3.4-4.4 g per column) in bare soil, due to the larger amount of irrigation water in columns cultivated with E. stagnina. At the end of the experiment, the SS had decreased more in soil cultivated with E. stagnina (by 65-87%) than in bare soil (by 34-45%).

Dynamics of Salt Stock during the Intercropping Season
Periodic water ponding during the intercropping season decreased the soil SS in cracked soil in all columns ( Figure 6). Although all columns received the same amount of water, more salt was removed from soil cultivated with E. stagnina (17-21 g per column) than from bare soil (10-15 g per column). At the end of the intercropping season, ponding in soil cultivated with E. stagnina had removed 31-39% of the soil SS present at the beginning of the intercropping season (i.e., 30-36% of initial SS in the columns). In contrast, ponding in bare soil had removed 17-25% of the soil SS present at the beginning of the intercropping season (i.e., 18-26% of initial soil SS).

Dynamics of Salt Stock during the Intercropping Season
Periodic water ponding during the intercropping season decreased the soil SS in cracked soil in all columns ( Figure 6). Although all columns received the same amount of water, more salt was removed from soil cultivated with E. stagnina (17-21 g per column) than from bare soil (10-15 g per column). At the end of the intercropping season, ponding in soil cultivated with E. stagnina had removed 31-39% of the soil SS present at the beginning of the intercropping season (i.e., 30-36% of initial SS in the columns). In contrast, ponding in bare soil had removed 17-25% of the soil SS present at the beginning of the intercropping season (i.e., 18-26% of initial soil SS).

E. stagnina as a Crop for Effectively Decreasing the Salinity of Vertic Soils
This study shows that the fodder crop of E. stagnina reduces soil salinity of salinesodic Vertisols. Indeed, after 15 months of growth on soil columns under laboratory conditions, the SS decreased considerably in soils cultivated with E. stagnina (65-87%) while it only decreased by 34-45% in submerged bare soil. The same tendency was also observed with the EC reduction which was more pronounced in soil cultivated by E. stagnina than in submerged bare soil, particularly in deeper layers.
At the end of the experiment, the initially saline soils had become non-saline (ECe < 4 dS/m). These results confirm the ability of E. stagnina to decrease soil salinity, as reported by previous studies [12,15,32]. Ref. [15] reported similar percentages of soil desalinization

E. stagnina as a Crop for Effectively Decreasing the Salinity of Vertic Soils
This study shows that the fodder crop of E. stagnina reduces soil salinity of saline-sodic Vertisols. Indeed, after 15 months of growth on soil columns under laboratory conditions, the SS decreased considerably in soils cultivated with E. stagnina (65-87%) while it only decreased by 34-45% in submerged bare soil. The same tendency was also observed with the EC reduction which was more pronounced in soil cultivated by E. stagnina than in submerged bare soil, particularly in deeper layers.
At the end of the experiment, the initially saline soils had become non-saline (ECe < 4 dS/m). These results confirm the ability of E. stagnina to decrease soil salinity, as reported by previous studies [12,15,32]. Ref. [15] reported similar percentages of soil desalinization under field conditions, observing that E. stagnina grown in saline Vertisols reduced the SS by 72-81% in the topsoil (0-10 cm) after 8 months of field growth in the Niger River valley. This decrease was similar to that of [12], who observed that E. stagnina grown for two years on a saline-sodic clay soil in Egypt reduced the soil ECe significantly (from 27.6 to 4.3 dS/m) in the 0-15 cm soil layer. These two studies were conducted under field conditions, while our laboratory conditions allowed for measurement of ECe at three depths and with drainage at the bottom of the columns. Pot experiments of vertic soil planted with Lotus corniculatus without drainage also experienced a decrease in ECe (from 8.37 to 2.11 dS/m) after 2 months of growth [53]. Several other studies have shown the impacts of salt-tolerant plants or halophytes to reduce soil salinity, such as Suadea maritima, Sesuvium potulacastrum and Ipomoea pers-caprae by [9], Lonicera japonica by [54], or Atriplex aucheri, Suaeda salsa and Salicornia europea by [41].
The reduction in soil salinity by E. stagnina is explained mainly by (i) its root system improving soil structure with a higher soil macroporosity, (6-9%) in comparison to bare soil (3-4%) as reported by [35], and increasing salt leaching which is the dominant process and (ii) by salt accumulation in plant biomass [15,32,41]. The relative importance of these two processes has also been reported by [16].

Modeling of Vertisol Soil Salinity Phytoreclamation by HYDRUS-1D
The Hydrus-1D model allowed simulation of the general trends in the phytodesalinization processes of Vertisols by the E. stagnina crop. Indeed, the simulated results agreed with the measurements, highlighting a decrease of the soil ECw and SS in soil cultivated with E. stagnina. At the end of the experiment, simulations and measurements showed that ECw and SS were greater in bare soil than in soil cultivated with E. stagnina. However, the simulations were performed without considering geochemical processes (e.g., salt dissolution or precipitation), which may have biased model predictions. Although model predictions followed the general trends of salt accumulation measured in plant biomass, the model predicted less salt in plant biomass, particularly during the second cropping season. Based on the threshold in the simulated salt accumulation predicted at the beginning of the second cropping season ( Figure 5), we assume that the prediction of water uptake and consequently salt uptake was the main source of uncertainty. The salinity stress threshold in the [51] function was calibrated for a plant (ryegrass) that does not necessarily have the same agronomic characteristics as E. stagnina.
Refs. [27,28] used Hydrus-1D to simulate the effects of gypsum amendment on soil salinity dynamics in a saline-sodic clay soil. Their predictions agreed with the measurements, showing a decrease in soil ECe in saline-sodic clay soil after gypsum application. Recently, ref. [29] coupled Hydrus-1D with the UNSATCHEM module to simulate the Ca, Na, Mg and K contents of sodic Vertisols reclaimed with gypsum amendment and irrigated with water of varying quality. Unlike our study, these studies excluded plant effects from the simulations, particularly the salt accumulation in plant biomass. We simulated soil with a single-porosity model during the cropping seasons and a double-porosity model during the intercropping season, assuming that the hydraulic parameters changed over time but were constant during each season. However, these hydraulic parameters may change progressively, given the shrink-swell processes of Vertisols [25,26,55]. These variations in the hydraulic parameters of Vertisols may affect soil desalinization, but they remain difficult to assess in our modeling approach. Moreover, other simulation models have been developed to consider geochemical processes (e.g., salt dissolution and precipitation). For example, PHREEQ-C [56] and SWAP [57] simulate both the salt accumulation in plant biomass based on plant growth and salt leaching based on the improvement in soil structure due to plant root systems.

Role of the Intercropping Season in the Desalinization of Vertisols
Soil cultivated with E. stagnina cracks more than bare soil, highlighting the benefits of E. stagnina roots, as reported by many authors [58,59]. Moreover, the intercropping season promotes soil restructuring during soil cracking, due to shrinkage mechanisms during the drying phase. The periodic water ponding on cracked soil during the intercropping season induced water flow bypass and salt transfer between the soil matrix and the macroporosity induced by cracks. The total amount of salt removed during the intercropping season by water ponding represented 30-36% of the initial SS in soil cultivated with E. stagnina and 18-26% of the initial SS in bare soil. Several authors [15,[60][61][62] also reported these processes, explaining that during the intercropping season, salt in cracked clay soils moves from the inside to the outside of soil prisms (due to evaporation), precipitates on prism faces and then becomes easily dissolved and leached by irrigation. However, ref. [63] reported that water flow bypass in cracks limits salt dissolution in the soil matrix, questioning the effectiveness of soil desalinization by water ponding on cracked soil. Moreover, supplying water causes early crack closure (4-5 h) in Vertisols [61,64] and low saturated hydraulic conductivity (2.8 × 10 −8 m/s), which decreases water infiltration and salt leaching [4,15]. The financial cost of moving water during this period without crop production may also limit the feasibility of this technique, particularly under arid and semi-arid conditions.

Conclusions
This study showed that the E. stagnina crop reduced the soil salinity of Vertisols and the Hydrus-1D model allowed simulation of the general trends in the phytodesalinization processes of Vertisols. Indeed after 11 months of growth, simulated and measured results showed that the SS, which was initially the same in all columns, decreased by 65-87% in soil cultivated with E. stagnina and by 34-45% in ponded bare soil. However, the simulated salt accumulation in plant biomass was lower than that measured, with a pronounced difference from the beginning of the second cropping season. Our results suggested that planting E. stagnina is a promising option for reducing the soil salinity of saline-sodic Vertisols. A simplified modeling approach based on Hydrus-1D was adequate for predicting the general dynamics of soil salinity. The fodder crop of E. stagnina can improve irrigated salt-affected lands and improve cropping systems by the diversification of agricultural production of irrigated land.